Characterization of aerosol sources and optical properties in Siberia using airborne and spaceborne observations

: Airborne backscatter lidar at 532 nm and in-situ measurements of black carbon (BC), carbon monoxide excess above background ( ∆ CO), and aerosol size distribution were carried out over Siberia in July 2013 and June 2017 in order to sample several kinds of aerosol sources. Aerosol types are derived using the Lagrangian FLEXible PARTicle dispersion model (FLEXPART) simulations and satellite observations. Six aerosol types could be identiﬁed in this work: (i) dusty aerosol mixture, (ii) Ob valley gas ﬂaring emission, (iii) fresh forest ﬁre, (iv) aged forest ﬁre, (v) urban emissions over the Tomsk/Novosibirsk region (vi) long range transport of Northern China urban emission. The altitude range of aerosol layers is discussed for each aerosol type, showing transport above the boundary layer for long range transport of Northern China emissions or fresh forest ﬁre. Comparisons of aerosol optical properties, BC and ∆ CO are made between aged and fresh plumes for both the urban and forest ﬁre emissions. An increase of aerosol optical depth at 532 nm (AOD 532 ), aerosol particle size and ∆ CO is found for aged forest ﬁre plumes. Similar results are obtained when comparing the aged urban plume from Northern China with fresh urban emissions from Siberian cities. A ﬂight above gas ﬂaring emissions corresponds to the largest AOD 532 and provides a possible range of 50–60 sr for the lidar ratio of these aerosol plumes often encountered in Siberia. Black carbon concentrations are relatively higher for the ﬂaring plume (0.4–0.5 µ g.m − 3 ) than for the urban plume (0.2 µ g.m − 3 ). The largest BC concentrations are found for the fresh forest ﬁre plume. The aerosol type identiﬁcation and AOD 532 provided by CALIOP Version 4.2 data products in air masses with similar origin generally agree with the results obtained from our detailed analysis of the aerosol plume origins.


Introduction
Atmospheric aerosols play a very important role in many meteorological, radiative and chemical processes taking place in the atmosphere such as cloud formation, scattering and absorption of incident solar (short-wave) and thermal (long-wave) radiation from the Earth, as well as affect the air quality [1]. Due to the variety of the optical, microphysical and chemical properties of atmospheric aerosols, closely depending on the formation processes of particulate matter and their subsequent aging processes occurring in the atmosphere, and due to the poor knowledge of their spatio-temporal distribution, they have been identified by the Intergovernmental Panel on Climate Change (IPCC) as one of the main uncertainty sources when assessing radiative forcing and the climate change [2]. airborne lidar. The aircraft campaigns, the instruments and the methodology to identify the aerosol type and to process the lidar data are described in Sections 2 and 3. In Section 4, a selection of six case studies representative of the major aerosol sources observed in Siberia are presented and analyzed in term of the expected aerosol mixing or age, the vertical extent of the layers and the aerosol optical properties. A selection of CALIOP profiles in air masses with the same origin as those observed by the aircraft also provide additional information on the optical properties of aerosols when using the Version 4.20 CALIOP aerosol data products. The discussion in Section 5 will focus on the comparative analysis of the different case studies.  The flights were performed over (i) the major large Siberian cities (Novosibirsk, Tomsk, Krasnoyarsk, Yakutsk) (ii) the gas flaring fields of the Ob valley and the industrial city of Norilsk (iii) as well as over the Siberian taiga in order to track the long-range transport of emissions from wildfires and mid-latitude Eastern Asia. So the major aerosol sources could be included in this analysis. During these two airborne campaigns, downward looking backscatter lidar measurements were carried out but also in-situ measurements of trace gas concentrations and aerosol particle properties (size distribution, scattering properties). The aircraft performed various legs at low altitude (≈0.6 km) and between 4 and 8 km in order to sample different atmospheric layers. Only lidar data collected during flights above 4 km are used for the aerosol layer characterization.

Airborne Lidar System
The lidar system installed on board the Tu-134 aircraft is based on the aerosol lidar developed at IAO [26,27]. The lidar is included in the scientific equipment of the "Atmosfera" Common Use Center of IAO SB RAS. The transmitter module is based on a solid state Nd-YAG laser emitting 8 ns laser pulses at 1064 and 532 nm. The maximum output energy at 532 nm is 100 mJ with a repetition rate of 10 Hz and a beam divergence of 2.5 mrad. The optical receiver is a 150 mm diameter reception lens coupled with a 1 nm filter and two reception channels (co-and cross-polarization). The full geometrical overlap is obtained between 80 m and 150 m. In practice the first 200 m are not used to reduce the errors when estimating the overlap function correction in clear air region below the aircraft. The detection unit is composed of a photomultiplier coupled to an analog-to-digital converter (ADC) electronic system with a sampling rate range of 25-100 MHZ (i.e., a 1.5-6 m vertical resolution) and a resolution of 12 bits. The cross-polarization calibration is not sensitive enough to characterize the aerosol type and is mainly used to discriminate cloud and aerosol layers. The near infrared channel was not available during the aircraft campaigns. Attenuated backscatter signal in decimal logarithm >−2.3 with a signal above detection threshold in the depolarization channel is considered as cloud. The initial lidar data temporal averaging is 8 s and 1 min, respectively, for the 2013 and 2017 campaign. After cloud screening, data are averaged over 1-5 min to get a measurement range higher than 6 km.

In-Situ Measurements
In-situ measurements include trace gas and aerosol measurements. CO measurement is performed using a fully automated CO analyzer based on a commercial infrared absorption correlation gas analyzer (Model 48C, TEI Thermo Environment Instruments, Franklin, MA 02038, USA). The instrument is described in [28]. The accuracy is 5 ppb (5% CO) for a 30 s integration time (i.e., the response time of the instrument) and the detection limit is 10 ppb. CO 2 , CH 4 and H 2 O measurements were performed by means of the Picarro G2301-m gas concentration analyzer with a 5 s precision of 70 ppb for CO 2 and 0.5 ppb for CH 4 . Water correction software automatically reports dry gas mole fractions.
Equivalent black carbon (EBC) mass concentration are measured using an aethalometer based on light attenuation by particles after collection on a filter [29]. The wavelength range between 0.4 and 1.1 µm with a maximum near 0.9 µm. This instrument is sensitive to submicron particles. BC mass concentration (ρ BC ) in µg.m −3 is converted from light absorption measurement (ln I I 0 ) with the following relationship: ρ BC = 697 · C f · ln I I 0 , where C f is the correction factor between 0.5 and 1 taking into account the blackening of the filter. The EBC sensitivity is 0.01 µg.m −3 .
The particle size distribution is fully characterized using two different instruments. Ultra fine particles concentration in the diameter range from 3 to 200 nm are measured using a diffusional particle sizer (DPS) consisted of an 8-channel automated diffusion battery (synthetic screen ADB; designed by ICKC SB RAS, Novosibirsk; Ankilow et al. [30], Ankilov et al. [31,32]) coupled with a condensation particle counter (TSI CPC 3781). One scanning period of the DPS takes 80 s to derive size distribution in 20 size bins. Transmission efficiency for the airborne instrument is corrected for and is ≈0.997 in the 70-200 nm range and between 0.82 at 400 hPa and 0.89 at 1000 hPa for the 3-70 nm size range. All concentrations are reported at standard pressure and temperature (STP) conditions. Particle concentrations in 31 size bins in the range from 0.25 to 32 µm are measured using a GRIMM 1.109 optical particle counter (GRIMM Aerosol Technik GmbH & Co. KG, Ainring 83404 Germany).
Meteorological parameters such as temperature, humidity and wind vector are measured routinely on-board using HYCAL sensor model IH-3602-C of Honeywell Inc. Temperature and relative humidity accuracies are 0.5 • C and 7%, respectively.

CALIOP Dataset
The CALIOP data analysis in this paper uses the vertical profiles of Level 1 (L1) Version 4.1 (V4.1) total attenuated backscatter coefficient (TAB) at 333 m horizontal resolution for both the visible (532 nm) and IR (1064 nm) channels and the TAB uncertainties are calculated using the ancillary parameters also available in the V4.1 L1 data products, namely the noise scale factor, laser energy, RMS random noise, calibration coefficient and amplifier gain. Cloud layer position and attenuation, aerosol type and lidar ratio are available at 5 km horizontal resolution in the V4.2 L2 layer cloud and aerosol data products. All the CALIOP cloudy profiles are removed regardless of the cloud aerosol discrimination (CAD) score to ensure that no clouds are accounted for in the analysis of TAB. In this work the CALIOP TAB and lidar ratio are horizontally averaged in the FLEXPART footprint corresponding to the aircraft observation (generally of the order of 100 km). The Fernald forward inversion needed to calculate the AOD and the backscatter ratio vertical profile at 532 nm (ratio of the total backscatter including molecular and aerosol component over the molecular backscatter) is made from the top altitude of the highest aerosol layer [33]. The uncertainty of the lidar ratio have been taken from [24] when calculating the error on the retrieved total backscatter coefficient. The particle linear depolarization ratio at 532 nm is calculated using the ratio of the perpendicular and parallel component of TAB and the backscatter ratio. The particle color ratio is defined as the ratio of the aerosol backscatter coefficient at 1064 and 532 nm. Both the particle linear depolarization ratio and the particle color ratio are not calculated for backscatter ratio at 532 nm less than 1.5 to remove noisy values.

Aircraft Data Processing
The lidar calibration is performed several times during each flight using a normalization of the attenuated backscatter signal (PR2) to molecular backscatter in the range 200-700 m below the aircraft with a 1-5 min temporal resolution. Flight sections without aerosol/cloud occurrence are determined using the in-situ aircraft measurements (total aerosol concentrations from the Grimm instrument <15 particles.cm −3 ). The vertical profile of molecular backscatter was estimated from the 0.75 • ERA-Interim ECMWF meteorological analysis [34]. To account for the the overlap function between the field of view of the laser and the telescope, the mean altitude dependency of the ratio between PR2 and the molecular backscatter is determined for each flight in the 0-700 m altitude range below the aircraft using only profiles with no cloud and aerosol layers. The calibration accuracy is then of the order of 5-10% due to the mean signal statistical uncertainty (<3%) and the assumption on the reference backscatter ratio being unity in the calibration range (≈5%).
The aerosol optical depth and the backscatter ratio vertical profile retrieval are based on the Fernald forward inversion of the calibrated PR2 [33], assuming a range independent value of aerosol lidar ratio (LR). The LR value is chosen according to the aerosol type identified following the methodology described in Section 3.2. This is somewhat similar to the approach proposed by [23] for the analysis of the CALIPSO aerosol extinction profiles. The range of the LR is taken from several papers providing measured LR either with a HSRL lidar or a Raman lidar for a large range of aerosol types. This will be discussed for each case study described in this paper.
To estimate the uncertainty of the retrieved backscatter ratio, 500 inversions were performed using random LR values within the possible LR associated to an aerosol type and random backscatter coefficient at the reference altitude within the interval corresponding to the statistical error on the attenuated backscatter at the reference altitude (≈5%). The complete methodology for the lidar data analysis is summarized in the upper panel of Figure 2.

Identification of the Aerosol Sources
Aerosol types of the layers observed by the airborne lidar were characterized using first the Lagrangian FLEXible PARTicle dispersion model (FLEXPART). FLEXPART is a Lagrangian model designed for computing the long-range transport, diffusion, dry and wet deposition, of air pollutants or aerosol particles backward or forward from point sources using a large number of particles [35,36]. For our study particle dispersion calculations are performed by including aerosol tracer removal processes by dry and wet deposition in the cloud and under the cloud. For each aerosol layer identified in a lidar profile, 5-10 days backward simulations of the spatial distribution of 10,000 particles released in a 1 km thick altitude zone is made. The occurrence of clouds is calculated by FLEXPART using the relative humidity fields. The meteorological fields used for the simulations (including precipitation rates) are operational ECMWF field at T255 horizontal resolution (≈80 km) and with 153 model vertical levels. FLEXPART simulations provide maps of Potential Emission Sensitivity (PES) for each aerosol layer observed by the airborne lidar. These maps represent the areas that have most influenced the observations including the correction due to the losses of the tracer [37]. The PES maps are always compared to the AOD 550 maps at 550 nm from the Level-3 MODIS Atmosphere Daily Global Gridded Product (1 • × 1 • resolution) in order to identify the effective aerosol source regions. Only data with high quality flag have been considered in this work.
The sources of biomass-burning aerosol are identified using the daily fire radiative power (FRP) maps based on NASA Fire Information for Resource Management System (FIRMS) from MODIS observations [38] and the Visible Infrared Imaging Radiometer Suite (VIIRS) [39]. Significant biomass burning aerosol production is taken into account only if the daily FRP is higher than 0.3 GW and the fire lifetime higher than three days and if the CO tropospheric column measured by IASI (Infrared Atmospheric Sounding Interferometer) is higher than the monthly background CO column [40][41][42] (Data set Clerbaux [43]). No quality flag is used, but cloudy pixel have been removed.
Dust outbreaks from the Eastern Asia desert are only taken into account if the aerosol optical depth measured by IASI at 10 µm (AOD 10µm ) is greater than 0.08 in the area with elevated PES [44][45][46] and when CALIOP aerosol depolarization ratio is greater than 15% in the same area [16,47,48]. No quality flag is applied to this data set.
The location of flaring sources is based on the anthropogenic emissions ECLIPSEv4 data-set (Evaluating the Climate and Air Quality Impacts of Short-Lived pollutants) described in [49]. This inventory includes in particular the gridded methane emissions from gas flaring in the Russian Arctic at a 0.5 • × 0.5 • horizontal resolution. A threshold of 50 moles.km −2 /h has been applied to the methane emissions to select areas that could potentially be defined as flaring sources. Urban pollution aerosol sources are considered only when large cities (>500,000 inhabitants) are included in the area with high PES (≥2500 s). The elevated CO tropospheric column (>2.10 18 molecule.cm −2 ) in the same area is also mandatory to select cities with significant industrial and urban combustion sources [50].
In addition to this first guess for the aerosol type identification based on the FLEXPART simulation analysis, in-situ aircraft measurements of CO concentration, black carbon (BC) mass concentration and aerosol size distribution are also analyzed for aircraft ascent or descent across the aerosol layer observed by the airborne lidar. Excess of CO (∆CO), i.e., the difference with the background CO concentration taken as the minimum of CO measured during the two campaigns in the lower troposphere (0-5 km), must reach 30 ppb for biomass burning aerosol and gas flaring emission (Paris et al. [4]). Black carbon mass concentrations is also used to identify combustion aerosol: BC > 0.5 µg.m −3 and BC maximum correlated with elevated ∆CO. The ratio of the aerosol concentration in the cloud condensation nuclei size range between 80 and 200 nm (N 80−200 ) over the aerosol concentration in the Aitken mode (N A ) between 15 and 80 nm is used to identify the aerosol aging [51][52][53]. The complete methodology for the aerosol type identification is summarized in the lower panel of Figure 2.

Aircraft Campaign Data Analysis
Six flights have been selected during the 2013 and 2017 campaigns, because they correspond to different aerosol sources. The lidar vertical cross-section of the calibrated attenuated backscatter (PR2) have been used to identify the horizontal and vertical extent of the aerosol layers (532 nm backscatter ratio larger than 1.5). For each aerosol plume, the aerosol type is retrieved using the methodology described in Section 3.2 and the AOD 532 of the layer is calculated with a constant LR in the layer as explained in Section 3.1.    The role of dust emission is also significant for this case study when looking at the 4-day average of the AOD 10µm measured by IASI which is in the range 0.15-0.35 in region 2 above Kazakhstan ( Figure 5a). A CALIOP overpass on 16 June 2017 at 21 UT also shows high particle linear depolarization ratio (15-20%) and a particle color ratio (ratio of aerosol backscatter coefficient at 1064 and 532 nm) in the range 0.3-0.7 in the lower troposphere at 50 • N, 70 • E in region 2 ( Figure 5b). An aerosol depolarization ratio less than 20% is consistent with polluted dust aerosol (Tesche et al. [47], Groß et al. [48], Burton et al. [15], Groß et al. [16]). In the flight area, IASI also detected a AOD 10µm between 0.08-0.1, i.e., smaller than above Kazakhstan, but being just above the dust detection threshold of 0.08. The aerosol layer observed by the airborne lidar at 57 • N, 80 • E can be therefore considered as a mixture of dust, industrial pollution the Ob Valley and aged smoke.

Dusty Aerosol Mixture
When looking at the in-situ measurements made by the aircraft during the descent at 57.5 • N, ∆CO range is between 20 and 30 ppbv up to 2.5 km and the BC mass concentration increases from 0.2 µg.m −3 at 2 km to 0.4 µg.m −3 at 500 m ( Figure 3b). These moderate values of ∆CO and BC compared to other flights (Table 1) are also consistent with a dusty aerosol mixture.
The AOD 532 and the backscatter ratio are calculated from the airborne lidar data for a 6 min average profile shown by the red rectangle in Figure 3. The lidar ratio range is between 29 and 49 sr according to [14,15] for a dusty mix aerosol type when using HSRL lidar measurements. Similar values (38 ± 7 sr) are reported by [54] with a Raman lidar in China for a dust plume mixed with urban emissions. The corresponding airborne lidar AOD 532 is then 0.1 ± 0.03 (Table 1). The average AOD 550 measured by MODIS within ±1 h and ±100 km around the lidar observations is similar (0.11) and shows very small variations (0.02). Looking at the FLEXPART PES analysis a CALIOP granule on 15 June 2017, 21UT also crosses the same air masses with the dusty aerosol mixture. The 80 km horizontal average of the backscatter ratio vertical profile at 44 • N is calculated using a Fernald inversion and the CALIOP lidar ratio available in the CALIOP Version 4.2 aerosol data products. It shows an aerosol layer up to 2.5 km but the backscatter ratio is smaller than the aircraft data between 1.5 and 2.5 km and shows a sharp peak up to 2.4 at 1 km ( Figure 6a). The CALIOP particle color ratio ≥ 0.5 (Figure 6b) is in agreement with a significant fraction of coarse size aerosol [55]. The CALIOP aerosol layer classification is indeed 77% polluted dust, 6% dust, 6% clean continental and 11% elevated smoke, with an average lidar ratio of 55.8 sr and a low average AOD 532 of the order of 0.11 ± 0.04.

Ob Valley Gas Flaring Emissions
On 18 June 2017, aerosol layers in the 0-3.5 km altitude range have been observed at higher latitudes during a second flight between Novosibirsk and Surgut above gas and oil extraction fields in the latitude range 58 • -59 • N. The layer is still very well identified as the aircraft approaches Surgut and in-situ measurements during the descent to Surgut have been made in the same aerosol layer (Figure 7).
Potential emission sensitivity higher than 2500 s and coupled with daily MODIS AOD 550 higher than 0.   Few specific studies of aerosol plumes from flaring zones with a Raman or HSRL lidar system have been published for flaring emissions. Chazette et al. [56] reported LR of the order of 71 sr at 355 nm at Hammerfest (70.6 • N, 23.6 • N), Northern Norway, for a mixture of flaring emissions from a single industrial source and aerosol transported from the Murmansk region (68 • N, 33 • E) in Russia. However, this case is hardly comparable to the case of the Ob valley, which has many flaring sources. The LR range of 40-50 sr, measured in Tomsk by [11] above the boundary layer between 2.5 and 5 km altitude using measurements with a Raman lidar at 532 nm in April and May 2007, is probably a better indicator of the range of LR for the flaring emissions of the Ob valley. The April-May 2007 period was not impacted by fires in Siberia and long distance transport was mainly controlled at 500 hPa by a westerly or northerly flow from the Ob valley region (22 days out of 31). The lidar ratio range is then chosen between 40 and 50 sr in order to calculate the AOD 532 and the backscatter ratio from the airborne lidar data for a 6 min average profile shown by the red rectangle in Figure 7a. The corresponding AOD 532 for the airborne lidar is then 0.21 ± 0.06 (Table 1), two times larger than the previous dusty mix case observed at 57 • N two days earlier. The average AOD 550 measured by MODIS within ±6 h and ±100 km around the lidar observations is also larger (0.33 ± 0.048). For a time slot of ±24 h around the aircraft measurement time, MODIS AOD 550 distribution is of the order of 0.23 ± 0.07. Regarding the CALIOP observations shown in Figure 10, 55% of the layers are classified as "Polluted Dust" and 45% as "Elevated Smoke" with a mean lidar ratio of 60 sr. The corresponding AOD 532 (0.32 ± 0.09) is very close to the MODIS observation at the same wavelength (0.35 ± 0.05) assuming an Angström exponent of 2 for this polluted continental plume. Therefore, the upper boundary (50 sr) of the LR range assumed for gas flaring is more likely.     To calculate the backscatter ratio and the extinction vertical profile, we assume a lidar ratio range of 33 sr to 46 sr based on the paper of [14] who provide LR for both fresh and aged biomass burning plumes in Northern America. This is also in good agreement with the 45 sr lidar ratio at 532 nm reported by [57] for a fresh fire plume in Southern Europe. The AOD 532 and backscatter ratio are calculated for three 40 s average profile shown by the red rectangles at 93.2 • E, 93.7 • E and 94.1 • E in Figure 11. The backscatter ratio reaches 4 in a 1 km thick layer for two sections and 3 for the other one. The AOD 532 are identical (0.118 ± 0.027) for layers (1) and (2) and less (0.085 ± 0.026) for the layer (3), knowing that the latter is underestimated by at least 35% since the layer may extend above 2.5 km when looking at the high concentrations measured by the aerosol counter at 2.8 km (200 particles/cm 3 ). The MODIS AOD 550 for ±1 h and ±100 km around the aircraft observations are significantly higher than the airborne lidar AOD 532 (0.19 ± 0.02). Such large differences are not surprising for spatially well defined aerosol plume close to the emission sources. Indeed MODIS AOD 550 becomes 0.155 ± 0.07 when the averaging time period is ±24 h. No CALIOP track was close enough to the aircraft to be compared with the same aerosol plume. However, a comparison was made with two CALIOP overpasses right above other fires with similar FRP (0.18 GW) located at 62 • N on 16 July 2013 at 19 UT and at 59 • N on 18 July 2013 at 05 UT ( Figure 13). The fires detected by CALIOP also show the altitude of backscatter ratio maximum in the range 2-3 km but with smaller scattering magnitude (of the order of 2.5). The color ratio range is 0.3-0.4 at the altitude of the fire plume, while the depolarization ratio is quite variable but remains below 10% ( Figure 13). The aerosol type identified by CALIOP are 58.5% and 100% polluted continental for, respectively the fires detected on 16 and 18 July 2013. No elevated smoke is detected because the layers remain below 3 km. The lidar ratio are taken as 59 and 70 sr for these two days, while the AOD 532 retrieved by CALIOP become 0.09 ± 0.04 and 0.17 ± 0.08. The CALIOP AOD 532 for fresh fires is in better agreement with the airborne lidar observations on 16 July 2013 probably because on this day CALIOP assumes a 59 sr lidar ratio more comparable to the range of LR for fresh fires.  (b) Vertical profiles of the particle color ratio (green) and particle linear depolarization ratio (red) measured by CALIOP.

Aged Forest Fires
It is also interesting to investigate data corresponding to aged fires which should provide larger extinction according to previous studies [15,47,58]. Part of the 18 June 2017 flight descending to Novosibirsk at 3 UT meets this objective. The latitudinal cross-section PR2, when the aircraft flies at 4 km, shows a layer of aerosol in the altitude range of 0 and 2 km with an horizontal extension of 120 km ( Figure 14).
The PES calculated by FLEXPART between 500 m and 2 km at 57.4 • N ( Figure 15) shows that aerosol emissions are from an area between 80 • E and 100 • E at 57 • N, while the highest MODIS AOD 550 in this area are related to forest fires which occurred 100 • E North of Irkutsk (0.05 < FRP < 0.2 GW). The CO tropospheric column measured by IASI North East of Irkutsk ( Figure 16) is of the order of 2.5 × 10 18 molecule.cm −2 , while it remains in the range 1.5-2.0 × 10 18 molecule.cm −2 above the cities of Tomsk, Novosibirsk and Krasnoyarsk also located in the area with high PES (≥2500 s). Forest fire plumes from Eastern Siberia at 2000 km from the aircraft can explain the aerosol layers seen by the airborne lidar at 57.49 • N, 78.63 • E on 18 June 2017. In-situ aircraft measurements ( Figure 14) are also consistent with this hypothesis since ∆CO concentrations are even higher than in the fresh fire plume (45-55 ppbv) and the 0.58 N 80−200 /N A ratio is significantly higher than the small values encountered in the fresh fire. The BC mass concentration (0.4-0.5 µg.m −3 ) is however weaker than the fresh forest value due to the short lifetime (few days) of BC in the atmosphere [59,60].   To compute the backscatter ratio and the AOD 532 from the airborne lidar attenuated backscatter shown in Figure 14a, the lidar ratio is taken between 55 and 73 sr, according to several studies in forest fire plumes, e.g., [61,62] reported lidar ratio of the order of 65 sr for aged Siberian smoke layers transported over Japan and Korea, while [14] is using a range of 55 sr to 73 sr for aged forest fire plumes. The backscatter ratio of the order of 2.3 is well mixed in the boundary layer but the upper altitude (2 km) is lower than for the fresh fire (3 km). The corresponding AOD 532 for the airborne lidar is 0.157 ± 0.03 (Table 1) so not significantly higher from those observed for the fresh fire. The MODIS AOD 550 for ±6 h and ±100 km around the aircraft observations are significantly higher than the airborne lidar AOD 532 (0.225 ± 0.044). Using MODIS AOD 550 for ±24 h around the aircraft flight does not change the results (0.21 ± 0.044). A MODIS AOD 550 daily variability for aged fire plume smaller than for aerosol plumes close to the source emission is expected with an increase in mixing during the smoke plume transport. Two CALIOP overpasses cross the high PES region between the aircraft (79  (Figure 17), i.e., farther away from the fire region. An increase of the color ratio with the lifetime of the fire plume is also consistent with the ratio N 80−200 /N A observed by the aircraft being higher than the ratio measured in the fresh fire case. The lidar ratio (65 sr) and the AOD 532 (0.15 ± 0.06) retrieved by CALIOP for these two layers are in good agreement with the results obtained with the airborne lidar. To reconcile the airborne lidar AOD 532 and CALIOP AOD 532 with MODIS AOD 550 is only possible if one assumes LR as high as 73 sr for these aged fires.

Siberian Urban and Industrial Emissions
The aircraft flew over the major Siberian cities in July 2013 between Krasnoyarsk and Novosibirsk, many aerosol layers have been encountered in the 0-3 km altitude range during this flight. On example has been selected on 20 July 2013 at 6 UT and is shown in Figure 18 when the aircraft is between Tomsk and Novosibirsk, i.e., a densely populated area with numerous industrial infrastructure. The PR2 latitudinal cross-section when the aircraft is flying at 4.5 km and then descends to 2.5 km, shows a 60 km long aerosol layer in the 0-3 km altitude range.   The range of LR values for pollution aerosol is quite large in published papers depending on the type of anthropogenic emission, on the season and on the altitude of the aerosol layer. For example, in Europe, Müller et al. [25] obtained 45-60 sr in Leipzig, while Burton et al. [15] proposed a LR range between 53 and 70 sr in North America using their airborne lidar flights around major North American cities and Mexico. In Beijing, Xie et al. [63] found 44 sr for high humidity and heavy pollution and 61 sr for lower humidity and moderate pollution level. Ansmann et al. [64] and Heese et al. [65] also found LR generally below 50 sr for the Pearl River delta in China where high pollution and humidity are generally found. At 355 nm, Dieudonné et al. [66] found LR (>90 sr) around several cities in Russia higher than the values found by Chazette et al. [67] in Paris (60-77 sr). For this study the LR range is then chosen between 43 and 90 sr to calculate the AOD 532 and backscatter ratio vertical profile for the 1 min average profile shown by the red rectangle in Figure 18. The AOD 532 for the airborne lidar is then 0.154 ± 0.088 (Table 1), while the backscatter ratio decrease with altitude from 2.3 to 1.5. The MODIS AOD 550 for ±1 h and ±100 km around the aircraft observations are quite high (0.29 ± 0.054) and remains high but with an elevated standard deviation for a distribution calculated using data ±24 h around the aircraft observations (0.24 ± 0.15). A CALIOP track is available on 17 July 2013 at 20 UT in the same area at 100 km east of the aircraft position. The CALIOP scattering ratio profile shows two layers with values higher than 1.6 below 1.3 km and at 2.3 km (Figure 21). The maximum value at 1.5 km is similar to the airborne lidar backscatter ratio. The CALIOP color ratio is between 0.3 and 0.5 in the layer at 2.3 km but increase significantly up to 0.7 below 1 km (Figure 21). This implies a significant contribution of the coarse mode aerosol compared to the case of flaring emissions. The CALIOP AOD 532 is 0.18 ± 0.06 with a lidar ratio of 70 sr, i.e., an AOD 532 similar to the aircraft observation. As for the fresh fire case, it is difficult to conclude about the MODIS AOD 550 being larger than the airborne or spaceborne lidar AOD 532 nearby the aerosol emission source even though the measurements are separated by less than 1 h and 100 km.

Long Range Transport of Northern China Emissions
The last case study corresponds to the lidar observations near the city of Yakutsk (62 • N, 129 • E) on 19 July 2013 at 11 UT. A PR2 vertical cross-section when slowly ascending west of Yakutsk shows several aerosol layers in the 0 and 5 km altitude range (Figure 22a).
The PES calculated by FLEXPART for the aerosol layer above 2 km and the MODIS AOD 550 4-day map show that the aerosol source region is located in a North-South corridor at 128 • E extending southward to the city of Harbin in Northern China (Figure 23). A smaller PES threshold of 2000 s has been chosen to define the source region for this layer because of significant removal of the aerosol tracer by washout during the transport from Northern China. The CO tropospheric column measured by IASI is also high (≥2.2 × 10 18 molecule.cm −2 ) in the PES maximum at 45 • N, 128 • E ( Figure 24). No fire is detected by FIRMS and no dust layer is seen by the AOD 10µm measured by IASI in the area between Harbin and Yakutsk ( Figure 24). Therefore, the main source of the aerosol layers observed by the lidar, and in particular the one located at altitude between 2.5 and 4 km, can be attributed to emissions from the Harbin region.   The AOD 532 and backscatter ratio profile are calculated for a 40 s average profile shown by the red rectangle in Figure 22 when the aircraft is high enough to sample the entire aerosol layer between 0 and 4 km. We could keep the same lidar ratio range considered in Section 4.5 for aerosol from Siberian cities, i.e., 43-90 sr. The backscatter ratio is again of the order of 2.2 in the layer near 3 km where the large ∆CO is measured. The AOD 532 for the airborne lidar is then 0.19 ± 0.1 ( Table 1). Because of the cloud cover in the southerly flow from Northern China on 17 and 18 July 2013, the number of MODIS AOD 550 measurements is very small and only MODIS AOD 550 average for a ±24 h time period is available (0.127 ± 0.04). The low AOD 550 compared to the summer AOD 550 values >0.4 usually encountered in Northern China [68] can be explained by the efficient aerosol removal in the southerly flow as shown also by the low BC concentrations measured by the aircraft. A CALIOP overpass on 18 July 2013 at 18 UT is perfectly located above the axis between Harbin and Yakutsk. A backscatter ratio vertical profile can be derived at 57 • N 12 h before the aircraft flight in Yakutsk and also shows an aerosol layer at 2.5 km above the PBL with a backscatter ratio of 2.3 similar to the measurements of the airborne lidar ( Figure 25). The CALIOP depolarization ratio is small (<5%) and the color ratio is of the order of 0.6 in this layer (Figure 25), i.e., as high as the value obtained in the polluted plume from the Siberian cities ( Figure 21) but higher than the color ratio measured above the gas flaring area (Figure 10). High color ratio is expected for this case study considering the aerosol aging and the increase of the coarse size fraction during the long range transport of polluted aerosol from Northern China [69]. The CALIOP aerosol classification is in good agreement with the FLEXPART analysis with 68% of "Polluted Continental" layers and 32% of elevated smoke, while the corresponding LR (70 sr) and AOD 532 (0.2 ± 0.07) is close to the aircraft observations 12 h later above Yakutsk.

Discussion
The characteristics of the six aerosol layers studied in this work are grouped in a summary table (Table 1). The dust/smoke layer transported from Kazakhstan and the aged smoke from Siberian fires have the lowest vertical extent (<2 km) with a nearly constant backscatter ratio of the order of 2. The limited vertical extent of the aged fire plumes have been already discussed in [13] using 2 years of lidar observations in Tomsk. On the contrary the vertical structure of fresh smoke plume shows a top altitude above 3.5 km and a well defined layer in the altitude range 2-4 km. The structure of urban plume from Northern China also shows a well defined layer in the same altitude range which can be explained by the upward transport in the frontal system along the West Pacific Coast. The two cases corresponding to urban and gas flaring plumes observed close to the emission area have a vertical extent of the order of 3 km and stronger vertical mixing between the ground and top altitude than for the fresh smoke and the urban plume transported from China. The meridional position of the plume east of 110 • E in the synoptic southerly flow from Northern China and the proximity of a fresh fire are then the main factors to explain the formation of well defined aerosol layers above 2 km above Siberia. The maximum backscatter ratio (3.5) is found for the fresh fire plumes while it is of the order of 2 for all other cases. Using either lidar AOD 532 (CALIOP and aircraft) or MODIS AOD 550 the smallest aerosol extinction (AOD 532 < 0.15) were obtained for the dusty mix and fresh smoke, while the largest (AOD 532 ≥ 0.2) are for the measurements on one hand above the gas flaring area in accordance with several studies on the impact of these emissions on aerosol production [5,70], and on the other hand for the long range transport of pollution [68]. It is worth mentioning that despite the large backscatter ratio for the fresh fire plume, the AOD 532 remains below 0.15. Increase of the lidar ratio at 532 nm due to aging of the aerosol and mixing of many individual fire plumes explain the 40% increase of the AOD 532 for the aged fire plume [14,57]. The largest differences between MODIS and lidar AOD are found for fresh plumes nearby the emission area (fire or urban pollution) where it becomes very important to take into account the measurement location mismatches. As explained in Section 4.2, the good agreement found between CALIOP and MODIS for the gas flaring case means that the lidar ratio range is probably 50-60 sr for the gas flaring plume sampled above Siberia.
Looking at the in-situ aircraft measurements, ∆CO increases for aged aerosol layers due to the long range transport of a biomass burning plume from Siberian fires or a polluted plume from Northern China. It is consistent with the CO lifetime being longer than 15 days and the buildup with time of individual plumes during the transport at regional scale [71,72]. BC mass concentration is high (>1.5 µg.m −3 ) for the fresh biomass burning plume while it is less than 0.5 µg.m −3 for the aged fire case because of efficient BC removal for transport lifetime > 1-2 days [59,60]. Gas flaring contribution to BC mass concentration (0.4-0.5 µg.m −3 ) is also higher than the urban pollution contribution either from local sources or transported from Northern China (0.2 µg.m −3 ). Moderate BC emissions from the local emissions of Siberian cities was not really expected considering the high value of ∆CO. It could be explained by a reduced combustion of wood and coal for urban heating during summer.
The lowest N 80−200 /N A ratios are obtained in the case of the fresh fire plumes (0.15) and the highest values in the case of the long distance transport of aerosol from China (0.8). In the case of aged fire plumes, local emissions from flares or urban pollution sources, the N 80−200 /N A ratio is between 0.3 and 0.6. Compared to the plume above the flares (ratio of 0.45 ± 0.15), aerosols in the aged fire plume (ratio of 0.6) probably have a significant fraction of terrigenous aerosol and exhibit an increase of the coarse size fraction due to coagulation during the transport from the source area [73,74]. The size spectrum measured by the DPS are shown in the supplementary document.
The color ratios from CALIOP observations are also consistent with the analysis of the N 80−200 /N A ratios. The lowest CR values (0.2-0.4) are obtained for the fresh smoke and flaring cases while the dust/biomass mixture and aged smoke cases show values between 0.5 and 0.7. Usefulness of color ratio to quantify the coarse to fine aerosol fraction has already been proposed by [55]. The aerosol depolarization ratio values, derived from CALIOP observations, are low for all cases (<5%) except for fire plumes where a terrigenous aerosol fraction again seems likely.
On the whole, the aerosol type classification from CALIOP products is consistent with the FLEXPART analysis except for cases of flaring and aged fires for which a significant contribution of desert dust is proposed but not justified by the detailed study of the air mass transport. The 20% negative difference between AOD MODIS and AOD CALIOP for the aged fire plume indeed shows that a lidar ratio associated with the "Elevated Smoke" (70 sr) rather than "Polluted dust" (55 sr) could explain such a difference.

Conclusions
Two airborne lidar campaigns were carried out over Siberia in July 2013 and June 2017. Aerosol types and optical properties were derived using FLEXPART aerosol tracer transport simulations and satellite data (MODIS, IASI and CALIOP). The aircraft in-situ measurements, mainly ∆CO and BC have been useful to validate the identification of the aerosol origin. Six aerosol types could be identified in this work: (i) Dusty aerosol mixture (ii) Ob valley gas flaring emission (iii) fresh forest fire (iv) aged forest fire (v) urban emissions over the Tomsk/Novosibirsk region (vi) long range transport of Northern China urban emission.
Aerosol plumes transported above the boundary layer have been only observed for the synoptic uplifting of Northern China emissions or vertical transport during forest fire plume. This study also allows comparisons between aged and fresh plumes for both the urban and forest fire emissions. The aged forest fire plume is characterized by an increase of AOD 532 and of the aerosol particle size in accordance with similar studies conducted in biomass burning aging in North America and Southern Europe. ∆CO increase (+10 ppbv) and BC decrease (≈−0.8 µg.m −3 ) have been also observed for the the aged fire plume, due to the accumulation of many individual plumes at regional scale and due to the BC wet or dry deposition. Similar results have been found for the comparison between the aerosol layers due to fresh and aged urban emission, especially the ∆CO increase (+20 ppbv) and the particle size increase (N 80−200 /N A ratio from 0.5 to 1 and CALIOP color ratio from 0.4 to 0.6).
The flight above gas flaring emissions has been useful to discuss the characteristics of these plumes which have not been very well studied by aircraft lidar observations. The largest AOD is found for this air mass type regardless of the kind of observations (aircraft, CALIOP, MODIS). A range of 50-60 sr is proposed for the lidar ratio in aerosol plumes related to Siberian gas flaring emission based on the good agreement between the AOD retrieved by CALIOP and MODIS. Black carbon concentrations are relatively higher for the flaring plume (0.4-0.5 µg.m −3 ) than for the urban plume (0.2 µg.m −3 ). Based on the analysis of the color ratio and the N 80−200 /N A ratio, the particle size in the flaring plume is slightly smaller than the aerosol size in the urban plume near Novosibirsk.
The aerosol type identification and AOD 532 provided by CALIOP Version 4.2 data products in air masses with similar origin generally agrees with the results obtained with our analysis of the aerosol plume origin, except for the flaring and aged fire layers where a significant contribution of dust is not justified. The CALIOP derived color ratio is also in good agreement with the behavior of the N 80−200 /N A ratio measured by the aircraft aerosol size distribution measurements. Such results are important considering the large amount of CALIOP observations available to characterize the optical properties over Siberia.
Author Contributions: A.Z. is the main contributor under the supervision of G.A., I.E.P., G.K., and Y.S.B. have been responsible of the airborne lidar measurements, while M.A., D.C. and V.K. made the aerosol, trace gas and BC measurements. J.P. provided the supervision of the satellite data analysis and J.-D.P. of the interpretation of the CO, aerosol size distribution data analysis. B.D.B. made possible the airborne campaigns. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded in France by Sorbonne Université and the French Centre National d'Etudes Spatiales (PhD contract 2921-2017). In Russia it is funded by a grant N • 075-15-2020-787 in the form of a subsidy for a major scientific project from Ministry of Science and Higher Education of Russia (project "Fundamentals, methods and technologies for digital monitoring and forecasting of the environmental situation on the Baikal natural territory").